Hartford HealthCare achieves 5% length-of-stay reduction with Epic-integrated H2O AI predictive analytics

Physician variability in assessing patient discharge readiness led to inconsistent length-of-stay outcomes, with less experienced physicians tending toward longer stays. Prolonged hospitalization beyond clinical necessity increased patient exposure to hospital-associated harms including infections, falls, and deconditioning.

Hartford HealthCare co-developed H2O (Holistic Hospital Operations) with MIT mathematician Dimitris Bertsimas — an AI/ML platform integrated directly into Epic that predicts patient discharge readiness within 24 hours of admission. The tool was embedded into multidisciplinary progression rounds, allowing care teams to compare physician-determined expected discharge dates against AI predictions to surface and resolve modifiable barriers.

Nearly 100% of admitted medical patients receive a discharge readiness prediction within 24 hours of admission. Combined with standardized progression rounds and geographic physician rounding, the health system achieved approximately a 5% reduction in overall length of stay compared to the pre-implementation period.

• Embedding physician leaders in the AI design process improved model accuracy and clinical adoption.
• Integrating the tool directly into Epic (rather than a standalone platform) significantly improved usability and workflow adoption.
• Combining predictive analytics with disciplined operational processes (standardized rounds, physician-unit alignment) multiplies the impact on throughput.